Download Reaction of Homopiperazine with Endogenous Formaldehyde: A

1521-009X/12/4008-1478–1486$25.00
DRUG METABOLISM AND DISPOSITION
Copyright © 2012 by The American Society for Pharmacology and Experimental Therapeutics
DMD 40:1478–1486, 2012
Vol. 40, No. 8
44917/3781111
Reaction of Homopiperazine with Endogenous Formaldehyde:
A Carbon Hydrogen Addition Metabolite/Product Identified in
Rat Urine and Blood
Scott Martin, Eva M. Lenz, Dave Temesi, Martin Wild, and Malcolm R. Clench
DMPK Department, Alderley Park, AstraZeneca UK Ltd., Macclesfield, Cheshire, United Kingdom (S.M., E.M.L., D.T., M.W.);
and Biomedical Research Centre, Sheffield Hallam University, Sheffield, United Kingdom (M.R.C.)
Received February 3, 2012; accepted May 1, 2012
ABSTRACT:
paign. These compounds were found to react with endogenous
formaldehyde from a rat in vivo study, resulting in the formation of
novel ⴙ13-Da bridged homopiperazine products (equivalent to the
addition of one carbon and one hydrogen atom), which were detected in urine and blood. The identification of these ⴙ13-Da products and their origin and mechanism of formation are described in
detail through analyses of a representative homopiperazine compound [N-(3-(3-fluorophenyl)-1,2,4-thiadiazol-5-yl)-4-(4-isopropyl1,4-diaze-pane-2-carbonyl)piperazine-1-carboxamide (AZX)] by
liquid chromatography-UV-mass spectrometry, 1H NMR, and
chemical tests.
Introduction
sophisticated MS subtraction routines such as mass defect filtering.
Experiments are therefore normally performed using state-of-the-art
instruments offering a variety of options to aid detection and structure
identification. The information is then used to direct and modify the
chemistry toward compounds with favorable metabolic properties.
Improved understanding of bioactivation mechanisms, reactive intermediate formation (so-called reactive metabolites), and adverse
toxicity has led to the front loading of biotransformation studies. In
response, most pharmaceutical companies now employ reactive metabolite trapping screens using liver microsomes (usually human)
fortified with nucleophiles such as GSH, cysteine, KCN, and methoxylamine (Prakash et al., 2008). The nucleophiles trap reactive
electrophilic species at sufficient concentration to favor the formation
of a stable product identified by their unique MS/MS spectral characteristics. Both early site of metabolism and reactive metabolite
trapping studies rely on in vitro systems to generate the metabolites.
However, metabolites can also be formed in complete biological
systems, which could be missed if the metabolic pathways are unknown and/or the endogenous reagents are not represented in these in
vitro systems. Hence, in our laboratories potential drug candidates
undergo nonradiolabeled in vivo metabolite identification studies.
These generally involve either a bile duct-cannulated study in rats,
with collection of urine and bile over a 24-h period or collection of
blood and urine from a high-dose rat pharmacokinetic study. The
studies aid the identification of the excreted metabolites and ensure
Understanding the metabolic fate of putative drug candidates both
in vitro and in vivo is a key component of drug discovery. Rapid
production of early information describing the rate of clearance and
site of metabolism is essential for directing iterative synthetic chemistry make-test cycles toward promising structural templates with the
requisite properties for an effective drug.
Reactive drug metabolites are of great concern in the pharmaceutical industry (Evans et al., 2004; Baillie 2006, 2009). Although their
identification is relatively straightforward with appropriate in vitro
trapping experiments, sometimes additional reactive compounds are
found unexpectedly. In general, early metabolism studies involve
incubation in hepatocytes or microsomes to mimic the most prevalent
metabolic processes occurring in the liver. Incubate samples at t ⫽ 0
min and at a terminal time point (usually 30 – 60 min) are then
compared by LC-UV-MS/MS. These studies can be both challenging
and time-consuming even when only a small number of metabolites
are identified. Simply mining the raw data to find the metabolites in
the terminal sample often requires the use of a variety of techniques,
ranging from simple UV comparison to complex common fragment
searching (commonly referred to as broad band or MSe) or the use of
Article, publication date, and citation information can be found at
http://dmd.aspetjournals.org.
http://dx.doi.org/10.1124/dmd.112.044917.
ABBREVIATIONS: LC, liquid chromatography; MS/MS, tandem mass spectrometry; MS, mass spectrometry; AZX, N-(3-(3-fluorophenyl)-1,2,4thiadiazol-5-yl)-4-(4-isopropyl-1,4-diaze-pane-2-carbonyl)piperazine-1-carboxamide; ACN, acetonitrile; MeOH, methanol; UPLC, ultra high-performance liquid chromatography; ESI, electrospray ionization; HCD, higher energy collisional dissociation; COSY, correlation spectroscopy;
ROESY, rotating frame Overhauser effect spectroscopy; ⫹ESI, positive ion electrospray ionization; RT, retention time.
1478
Downloaded from dmd.aspetjournals.org at ASPET Journals on June 17, 2017
Drug reactivity and bioactivation are of major concern to the development of potential drug candidates in the pharmaceutical industry (Chem Res Toxicol 17:3–16, 2004; Chem Res Toxicol 19:889–
893, 2006). Identifying potentially problematic compounds as soon
as possible in the discovery process is of great importance, so
often early in vitro screening is used to speed up attrition. Identification of reactive moieties is relatively straightforward with appropriate in vitro trapping experiments; however, on occasion unexpected reactive intermediates can be found later during more
detailed in vivo studies. Here, we present one such example involving a series of compounds from an early drug discovery cam-
REACTION OF HOMOPIPERAZINE WITH FORMALDEHYDE
Materials and Methods
Chemicals and Suppliers. Compound AZX was synthesized and developed at AstraZeneca UK Ltd. (Macclesfield, UK). Acetonitrile (ACN), meth-
anol (MeOH), ammonium acetate (analytical reagent grade), and formic acid
were acquired from Thermo Fisher Scientific (Loughborough, UK). Potassium
cyanide, formaldehyde (37% in H2O), and the deuterated NMR solvents were
sourced from Sigma-Aldrich (Poole, UK).
Standard/Stock Solutions. AZX stock solution. AZX was dissolved in
MeOH at a concentration of 200 ␮M.
AZX test solution. AZX was dissolved in MeOH/water (30:70%, v/v) to a
concentration of 10 ␮M. The test solution was prepared from the stock
solution. (Equivalent stock and test solutions were prepared in ACN to assess
whether MeOH was a contributing factor in AZX ⫹ 13 formation.)
Formaldehyde. Formaldehyde (37% in H2O) was used undiluted in all
spiking experiments.
KCN. KCN was prepared in water to a concentration of 50 mM.
Animal Dosing and Sample Collection (Urine and Blood). Dosing solution. AZX was prepared in a formulation of dimethylamine-water (40:60, v/v)
at a concentration of 1 mg/ml.
Rats. Male Han Wistar rats (n ⫽ 4) were divided into two groups (n ⫽
2/group), each receiving a single dose of AZX at 2 mg/kg i.v. at a dose volume
of 2 ml/kg. Urine was collected predose and at 0 to 6, 6 to 12, and 12 to 24 h
from group 1, whereas group 2 provided blood via the tail vein predose and at
20 min, 1.5 h, 6 h (at a volume of 0.3 ml), and finally 24 h (at a volume of 1.3
ml). Rat urine and blood were stored frozen at ⫺20°C until further analysis.
Additional predose/control rat urine samples were provided on request,
throughout the study.
Sample Preparation. Urine. Urine samples (0 –24 h postdose) were pooled
using a set volume (100-␮l aliquots) from each time point before analysis.
Predose urine (190- and 90-␮l aliquots) was spiked with 10 ␮l of the AZX
stock solution (200 ␮M) to give a final concentration of 10 and 20 ␮M,
respectively. Both the predose and 0 to 24 h urine samples were centrifuged at
approximately 12,000g for 5 min before analysis. The supernatant was then
FIG. 1. Proposed LTQ Orbitrap collision cell
fragmentation of compound AZX (top) with
accurate mass spectrum (bottom). MW, molecular weight; MF, molecular formula.
Downloaded from dmd.aspetjournals.org at ASPET Journals on June 17, 2017
identification of any unexpected reactive metabolites not generated or
detected in the preliminary in vitro systems.
During recent rat in vivo metabolite identification studies with a
series of lead compounds, cyclized GSH adducts were detected,
similar to those reported by Doss et al. (2005), highlighting a potential
reactive metabolite alert. This alert was not raised in the conventional
GSH trapping screen, because of the type of the GSH rearrangement
(data not shown). The reactophore in these lead compounds consisted
of a terminal piperazine, which was responsible for this bioactivation.
To preserve the potency of the compounds and remove the reactive
metabolite risk, the chemistry was changed to a homopiperazine
series.
Hence, several promising homopiperazine compounds from this
series were dosed to rats to assess whether these GSH adducts were
also formed. Instead, the analyses led to the observation of unusual,
novel products with a molecular weight gain of 13 Da (while showing
an apparent increase of ⫹12 Da by mass spectrometry) as the major
parent-related material in the urine and blood samples, which were not
detected in the preliminary in vitro studies. The formation of these
products, referred to as N-(3-(3-fluorophenyl)-1,2,4-thiadiazol-5-yl)-4(4-isopropyl-1,4-diaze-pane-2-carbonyl)piperazine-1-carboxamide
(AZX) ⫹ 13 throughout, is subject to further investigation in this
article, with compound AZX (Fig. 1) as a representative structure for
the “homopiperazine series.”
1479
1480
MARTIN ET AL.
12,376 Hz, resulting in an acquisition time of 2.64 s. A relaxation delay of 2.4 s
was used to ensure T1 relaxation between successive scans and, depending on
concentration, approximately 64 to 512 scans were acquired per sample.
Two-dimensional 1H-1H COSY (gradient enhanced; Bruker Biospin Ltd.,
Coventry, UK) experiments were used on the samples to determine signal
connectivities. Here, spectra were acquired into 4K data points in F2, and 128
increments in F1. The spectral width was set to 8012 Hz, resulting in an
acquisition time of 0.26 s. A relaxation delay of 1.5 s was used between
successive scans; 128 increments were acquired in F1, consisting of four scans
each. Before Fourier transformation, the data were apodized with a sine bell
window function, linearly predicted to 512 data points and zerofilled in F1 to
1024 data points. Selective excitation experiments were performed to confirm
the structure of the AZX ⫹ 13 ⫹ CN adduct.
A one-dimensional selective ROESY experiment (selrogp; Bruker Biospin
Ltd.) was performed. The data were collected into 65K data points, over a
spectral width of 12,019 Hz, resulting in an acquisition time of 2.77 s. A
relaxation delay of 2.4 s was used and a spin lock time of 100 ms.
Results
Structural Characterization of Compound AZX. Test compounds often contain structural motifs similar to those of their metabolites; it is therefore common practice to first fully characterize the
structure of the test/parent compound with accurate mass MS/MS
fragmentation. Characteristic fragment ions identified from the test
compound MS/MS can then be used to find and elucidate metabolite
structures.
Analysis of AZX by LC-UV-MS-MS/MS yielded a protonated
molecular ion with an accurate mass of [M ⫹ H]⫹ 476.2243 (⫹0.9
ppm). The proposed fragmentation pattern and LTQ Orbitrap HCD
MS/MS spectrum of compound AZX (Fig. 1) revealed two intense
key fragments, m/z 255.2175 (corresponding to the loss of the fluorophenyl-thiadiazol group of AZX) and m/z 141.1385 (corresponding
to the homopiperazine isopropyl moiety).
Detection of the Novel Metabolite with m/z 488 (AZX ⴙ 13) in
the In Vivo Samples. Typically, in early in vivo metabolite identification studies, temporal blood and urine samples are typically combined to obtain a single pooled sample for each matrix. The sample is
then analyzed on a LC-UV-MS/MS system in conjunction with a
predose sample spiked with analyte to obtain a final concentration of
10 ␮M (a concentration shown to produce a discernible UV response
in the biological matrix).
Spiking of the predose serves two purposes: it helps to rule out
synthetic impurities that could be misinterpreted as metabolites and to
discount endogenous material visually or by automated data subtraction.
In this study, only low levels of compound AZX were detected
upon analysis by UPLC-UV-MS/MS in both the 0 to 24 h rat urinepooled sample and the AZX-spiked predose urine sample. However,
an unexpected major parent-related peak was detected with a molecular ion of m/z 488. This observation was confirmed by respiking of
parent compound into samples of fresh predose/control urine (representing final AZX concentrations of 10 and 20 ␮M), which, once
again, resulted in the spontaneous formation of the m/z 488 product as
the major parent-related component, at an average ratio of approximately 95:5 (m/z 488: AZX parent) (data not shown).
It was initially assumed that either the wrong compound was dosed
or that the parent had either degraded/formed chemical adducts,
because the molecular ion, i.e., the mass addition, could not be
explained. Therefore, to verify the identity and stability of the parent,
the solvent standards (i.e., the 200 ␮M AZX stock solution and the 10
␮M AZX test solution) and the actual dose solution (diluted to 10 ␮M
in 30:70, methanol/water, v/v) were analyzed in conjunction with
repeat predose and 0 to 24 h rat urine samples. A comparison of the
UV chromatograms (extracted at ␭ ⫽ 240 –245 nm) for the AZX-
Downloaded from dmd.aspetjournals.org at ASPET Journals on June 17, 2017
transferred to Agilent high-performance liquid chromatography 2-ml vials with
200-␮l inserts for LC-MS analysis.
Blood. Blood samples (0 –24 h postdose) were pooled using a set volume
(50-␮l aliquots) from each time point. Blank/predose rat blood was spiked with
10 ␮l of the AZX stock solution (200 ␮M). Both the predose/blank and 0 to
24 h blood were diluted 1:1 with H2O and quenched with chilled (4°C) ACN
(1:3, v/v) followed by centrifugation at 12,000g for 5 min, and the supernatant
was transferred to Agilent high-performance liquid chromatography 2-ml vials
for LC-MS analysis.
Formaldehyde Addition. The formaldehyde-spiked AZX test solution was
prepared by adding 10 ␮l of formaldehyde (37% in water) to 500 ␮l of the
AZX test solution (10 ␮M, MeOH/H2O, 30:70, v/v) and analyzed by LC-MS
immediately.
Potassium Trapping Experiments. Ten-microliter aliquots of the KCN
solution (50 mM KCN/water) were added to a) 90 ␮l of the 0 to 24 h rat urine
pool, b) 90 ␮l of predose rat urine, which was spiked with 10 ␮l of AZX (200
␮M, i.e., the stock solution) to give a final concentration of 20 ␮M, and c) 90
␮l of the formaldehyde-spiked AZX test solution (detailed above). Both the
urine samples and the formaldehyde-spiked AZX test solution were left at
room temperature for 24 h before injection onto the LC-MS system.
Preparation of NMR Samples. Here, a more concentrated solution of AZX
was prepared, at a concentration of 1 mM in ACN. For NMR spectroscopy,
three solutions were prepared: a) 2 ml of AZX/ACN (1 mM); b) solution a with
100 ␮l of formaldehyde (37% in H2O); and c) 1 ml of solution b with 100 ␮l
of KCN (50 mM). Each of the solutions was stored for 24 h (to allow complete
formation of the products) before evaporation of the solvents. The solvents
were evaporated under nitrogen; each sample was freeze-dried and reconstituted in 200 ␮l of MeOH-d4 for 1H NMR spectroscopic analysis.
Identification and Structural Characterization of AZX ⴙ 13 by
UPLC-LTQ Orbitrap. Accurate mass structural characterization was acquired on a LTQ Orbitrap XL connected to a Waters Acquity UPLC system.
The Waters Acquity system consisted of a binary UPLC PUMP, a column oven,
an autoinjector, and a photodiode array detector. Separations were performed out
on a Kinetix C18 column (100 ⫻ 2.0 mm, 2.6 ␮m; Phenomenex, Macclesfield,
UK) preceded by a guard filter in an column oven at 50°C. The mobile phase
consisted of different solvent systems to assess the contribution of solvent/
buffer/acidifier to the AZX ⫹ 13 formation: 1) ammonium acetate (5 mM,
eluent A) and methanolic ammonium acetate (5 mM, eluent B) and 2)
formic acid (0.1%, eluent A) and formic acid/ACN (0.1%, eluent B).
The AZX stock and test solution were found to be stable in both mobile
phase systems. For the in vivo samples, the AZX ⫹ 13 product was detected
at approximately the same concentrations in either solvent system used. Hence,
solvent system 1 was subsequently used routinely.
The elution profile was linear gradient 90% A to 10% A, 0.00 to 8.00 min;
isocratic hold, 10% A 8.00 to 10.00 min; and reequilibration 90% A, 10.01 to
13.00 min. The flow rate was 0.6 ml/min, and the eluent was introduced into
the mass spectrometer via the LTQ divert valve at 1 min. The injection volume
was 20 ␮l, and UV spectra were acquired over 190 to 330 nm. The LTQ
Orbitrap XL was equipped with an electrospray ionization (ESI) source
(Thermo Fisher Scientific, Bremen, Germany), which was operated in positive
mode. Source settings were as follows: capillary temperature, 350°C; sheath
gas flow, 25; auxiliary gas flow, 17; sweep gas flow, 5; source voltage, 3.5 kV;
source current, 100.0 ␮A; capillary voltage, 18 V; and tube lens, 75.0 V.
Full-scan MS data were obtained over the mass range of 100 to 1000 Da at a
peak resolution of 7500. Targeted MS/MS experiments were acquired using
higher energy collisional dissociation (HCD) fragmentation, isolation width 2
Da, normalized collision energy 45, and activation time 30 ms. HCD fragment
ions were monitored by the Orbitrap using 7500 resolution. LTQ and Orbitrap
mass detectors were calibrated within 1 day of commencing the work using
ProteoMass LTQ/FT-Hybrid ESI positive mode calibration mix (Supelco,
Bellefonte, PA).
1
H NMR Spectroscopy for Structure Verification. 1H NMR analyses
were performed to confirm and structurally characterize AZX ⫹ 13 and the
AZX ⫹ 13 ⫹ CN adduct. 1H NMR data were acquired on a Bruker AVANCE
600 MHz spectrometer, operating at 600.13 MHz 1H resonance frequency. The
NMR spectrometer was equipped with a 2.5-mm SEI 1H/19F probe.
One-dimensional 1H NMR spectra of substrate and product were acquired
without solvent suppression into 65K data points over a spectral width of
REACTION OF HOMOPIPERAZINE WITH FORMALDEHYDE
The proposed fragmentation pattern and LTQ Orbitrap HCD
MS/MS spectrum of this metabolite/product (RT ⫽ approximately
4.45 min) (Fig. 3) contained diagnostic fragment ions m/z 267.2184
and m/z 153.1386. These ions corresponded to the addition of 12 Da
to the key fragments m/z 255 and m/z 141 in the MS/MS spectrum of
AZX. The observation of the fragment ions m/z 267.2184 and m/z
153.1386 with accurate mass (⫾2 ppm) and elemental composition
analysis appeared to confirm the addition of a single carbon atom to
the homopiperazine ring. The exact position of the carbon addition
could not be determined by MS; however, it was possible to postulate
the structure as a bridged homopiperazine (as shown in Fig. 3).
Difference between Theoretical Molecular Weight and Measured Molecular Ion of the Proposed Bridged Homopiperazine.
The proposed bridged structure equates to the addition of one carbon
and one hydrogen atom, i.e., a gain of 13 Da, which is inconsistent
with the 12-Da increase as determined by the mass spectrometry data.
However, because the bridged product has a fixed permanent positive
charge (M⫹), it cannot produce a protonated molecular ion [M ⫹ H]⫹
by positive ion electrospray (⫹ESI). Hence, whereas the calculated
nominal mass of the parent (AZX) is 475 and the mass measured by ⫹ESI
mass spectrometry ([M ⫹ H]⫹) is 476 (Fig. 1), the calculated nominal mass
and the mass measured by ⫹ESI MS of the bridged ion are both 488.
Therefore, a comparison of the theoretical molecular weights of AZX and
AZX ⫹ 13 results in a mass difference of 13 Da, compared with the 12-Da
difference in the measured molecular ions by mass spectrometry. To highlight the ⫹13 Da addition of this structurally unique product compared with
the ⫹12 carbon addition products reported in the literature (see Discussion),
the product is referred to as AZX ⫹ 13.
KCN Addition as a Chemical Test to Confirm the Presence of
a Quaternary Nitrogen. Cyanide chemically forms adducts with
iminium ions and is used widely in reactive metabolism trapping
FIG. 2. UV chromatograms of the AZX test solution (10 ␮M) (top)
compared with a predose rat urine sample after spiking with AZX
(10 ␮M) (bottom). ␮AU, micro-absorption units.
Downloaded from dmd.aspetjournals.org at ASPET Journals on June 17, 2017
spiked predose urine samples (at 10 and 20 ␮M final concentrations)
and the AZX solvent standards showed that in urine the UV peak for
AZX (RT ⫽ approximately 5.32 min) was depleted, whereas the m/z
488 product (RT ⫽ approximately 4.45 min) was abundant (Fig. 2).
However, the AZX solvent standards and the dose solution resulted in
the observation of the correct molecular ion of [M ⫹ H]⫹ 476,
confirming that the compound had not degraded in solution over time.
Similar observations were made with the 0 to 24 h rat blood, in
which the m/z 488 product was also observed, although at a reduced
amount (at an approximate ratio of 40:60, m/z 488: AZX) as assessed
by UV (data not shown). Furthermore, the spiking experiments with
predose blood, as conducted with the urine samples, again confirmed
the formation of the m/z 488 product. From these data it was surmised
that AZX was reacting with a component present in the urine and
blood, which is the subject of this investigation.
A thorough review of the 0 to 24 h sample data (urine and blood)
confirmed that this m/z 488 metabolite/product (RT ⫽ approximately
4.45 min), represented the majority of the AZX-related material in
urine as determined by UV and mass spectrometry. The mass of the
m/z 488 product was equivalent to an increase of 12 Da from parent
(AZX, [M ⫹ H]⫹ 476), although it represented an actual increase of
13 Da when molecular weights were compared (as detailed in the
following section).
Structural Characterization of the Novel Metabolite/Product,
m/z 488 (AZX ⴙ 13) by UPLC-MS/MS. From accurate mass measurement this apparent metabolite was determined to have a monoisotopic mass of 488.2242, almost exactly an increase of 12.0000 Da
over the parent [M ⫹ H]⫹ 476.2243, suggesting addition of one
carbon atom. This was further substantiated after an elemental composition analysis in which no rational alternative molecular formula
could be identified.
1481
1482
MARTIN ET AL.
Downloaded from dmd.aspetjournals.org at ASPET Journals on June 17, 2017
FIG. 3. Proposed LTQ Orbitrap collision cell fragmentation of the postulated bridged homopiperazine (AZX ⫹ 13) (top) with accurate spectrum (bottom). MW, molecular
weight; MF, molecular formula.
studies in biological samples across the pharmaceutical industry
(Argoti et al., 2005). To confirm the presence of a quaternary
nitrogen (i.e., the bridged homopiperazine moiety as shown in Fig.
3), KCN was added to the AZX-spiked predose and the pooled 0 to
24 h urine sample. On addition of KCN, the AZX ⫹ 13 product
peak (m/z 488, RT ⫽ approximately 4.45 min) reduced in size
(based on its UV response), whereas an additional peak [M ⫹ H]⫹
515 (RT ⫽ approximately 5.85) was detected in each of the
samples (Fig. 4). Further investigation by MS/MS fragmentation
determined the addition of 27 Da (CN) on the homopiperazine ring,
by the presence of the key fragment ions m/z 294.2289 and m/z
180.1497, corresponding to ⫹27 Da on m/z 267.2179 and m/z
153.1386, respectively. Confirmation of the presence of the quaternary nitrogen on the homopiperazine led to the investigation of
the mechanism of formation of AZX ⫹ 13.
Investigation of the Formation of AZX ⴙ 13. On the basis of the
MS results and the KCN trapping experiment (confirming the quaternary nitrogen), it appeared that AZX was reacting with a component
in the urine to produce a one carbon and one hydrogen atom addition
bridged homopiperazine.
REACTION OF HOMOPIPERAZINE WITH FORMALDEHYDE
The same product was observed with pooled blood, albeit to a lesser
extent. Formaldehyde, reported to occur naturally in living systems
(Heck and Casanova, 2004), was suggested as a likely candidate to
generate AZX ⫹ 13, via a quaternary Schiff base intermediate (iminium ion), which is then intramolecularly stabilized by forming the
bridged homopiperazine (as shown in Fig. 5).
To test this hypothesis, formaldehyde (10 ␮l, 37% in H2O) was spiked
into the AZX test solution (10 ␮M, 500 ␮l), which was analyzed
immediately on the UPLC-MS system. A product was formed at nearly
100% yield within the time taken to inject the sample, confirming that
formaldehyde reacts rapidly with AZX at room temperature. This chemical product was verified as the AZX ⫹ 13 product, which was
identical to that detected in the biological samples (urine and blood),
as assessed by chromatographic retention time, UV, accurate mass,
and MS/MS fragmentation. This now provided an efficient means to
generate AZX ⫹ 13 at a sufficient scale for full structural confirmation by 1H NMR spectroscopy, as detailed under Methods and Materials. Not only was the confirmation of the bridged homopiperazine
structure of importance, but the resultant structure of the AZX ⫹ 13
CN adduct was also of interest to confirm the exact position of the CN
addition, in view of several possible isomeric products.
Confirmation of AZX ⴙ 13 and the AZX ⴙ 13 ⴙ CN Adduct by
1
H NMR Spectroscopy. After identification by MS, the structures of
the parent (AZX), the AZX ⫹ 13 product, and the AZX ⫹ 13 ⫹ CN
adduct were verified by 1H NMR spectroscopy (Fig. 6, A, B, and C,
respectively). After the addition of formaldehyde, no changes in the
spectra were observed in the aromatic regions of AZX and AZX ⫹ 13
spectra, indicating that the site of modification is remote (data not
shown).
However, the aliphatic region highlighted several chemical shift
changes supporting the proposed formation of the bridged homopiperazine (Fig. 6, A and B). These mainly comprised changes in
the chemical shifts of the protons on the homopiperazine moiety, such
as the isopropyl methyl doublets, which have shifted from 0.94 and
0.98 to 1.34 and 1.38 ppm, respectively. Likewise, all the residual
OH
O
N
R1
H
Discussion
This study demonstrated that compound AZX and various analogs
containing a homopiperazine moiety reacted rapidly with formaldehyde in biological matrices to form a carbon and hydrogen addition
bridged homopiperazine (⫹13-Da products). However, the origin/
source of formaldehyde was still unclear, because it could have been
endogenous, as reported by Heck and Casanova (2004) or, indeed,
derived from the solvent system (methanol or formic acid), although
it was shown to form irrespective of the mobile phase combination or
spiking solvent used. A series of publications have described the
formation of a ⫹12-Da product from a reaction with formaldehyde,
which was suggested to have derived from methanol, as discussed
below.
It has to be pointed out that in these articles the reaction with
formaldehyde resulted in ⫹12-Da heterocyclic products (having
gained one carbon atom over their respective parent compounds),
whereas our example, through a different mechanism, yielded a novel
bridged homopiperazine with a quaternary nitrogen (after addition of
one carbon and one hydrogen). Hence, Yin et al. (2001) described the
addition of ⫹12 Da to the parent compound identified in/derived from
in vitro experiments, in which S9 fractions/hepatic incubations were
spiked with an analyte in methanol. However, it was concluded that
the MeOH was first metabolized to formaldehyde (Teschke et al.,
1974), which then reacted with a basic group on the incubated
compound. Each of the compounds tested, containing either 1,2amino hydroxyl or 1,2-diamino reacted with the metabolic formalde-
H
CH2
N
N
R1
N
N
+
R1
Schiff base
i t
intermediate
di t
H
+
N
N
H
R1
FIG. 5. The proposed structure and formation
of the bridged homopiperazine (AZX ⫹ 13)
through a quaternary Schiff base intermediate.
Downloaded from dmd.aspetjournals.org at ASPET Journals on June 17, 2017
homopiperazine protons experienced a shift to higher frequency. In
addition to the chemical shift changes, two doublets (labeled Ha and
Hb), not initially present in the parent spectrum, were observed at 4.68
and 3.91 ppm, with a coupling constant of 3J ⫽ 9.82 Hz and an
integral value of one proton per doublet, which showed a clear
connectivity in the two-dimensional COSY spectrum (data not
shown). The suggested structure is consistent with the proposed one
carbon bridge across the homopiperazine.
Figure 6C shows the 1H NMR spectrum of the AZX ⫹ 13 ⫹ CN
adduct. The ⫺CN addition was shown to have largely reversed the
chemical shift changes induced by the addition of formaldehyde (the
formation of the bridged homopiperazine), yet an additional set of
doublets (an AB system, labeled Hc and Hd) was noted at 3.7 and 3.5
ppm, which showed a strong connectivity in the two-dimensional
COSY spectrum (data not shown). On the basis of this evidence, two
isomeric possibilities could be suggested (Fig. 7, isomers 1 and 2).
Selective irradiation experiments were used to determine the correct
structure of the AZX ⫹ 13 ⫹ CN adduct.
The selective ROESY experiment, irradiating the isopropyl methyl
signals (labeled 8,8⬘), showed clear through-space interactions to the
isopropyl proton (labeled 9 in the structure), as well as the homopiperazine protons 2 and 3 (Fig. 7). There was no enhancement of
the bridged protons (labeled a and b), indicative of the six-membered
ring (isomer 2), hence providing evidence that this was not the
preferred structure. Instead, the spectral data supported the presence
of the NCH2CN side chain (isomer 1), after addition of KCN.
FIG. 4. The UV chromatogram of the predose urine after addition of KCN. ␮AU,
micro-absorption units.
N
1483
1484
MARTIN ET AL.
R
6
N
O
N
7
8,8’
O
N
1 2
N
6eq, 6ax
7eq, 7ax
1
5
2eq
2ax 3ax
5ax
3eq
5eq
8
9
N
4
A
8’
3
4eq
4ax
9
R
N
O
Hb
Ha
2eq
N
N
9
3ax
3eq
2ax
N
a, b
4ax
N
B
+
4eq
5eq
5ax
8,8’
Hc
1
C
Hd
5, 5’
5
9
2eq
2ax
3,3’
4,4’
4,4
FIG. 6. 1H NMR spectra of (A) the parent (AZX), (B) the AZX ⫹ 13 product, and (C) the CN adduct (AZX ⫹ 13 ⫹ CN).
hyde to generate ring-closed heterocyclic or ⫹12-Da products. Cunningham et al. (1990) reported a similar problem with diaminotoluene
reacting with formaldehyde generated from the use of methanol as a
spiking solvent for in vitro incubations. The product was formed
through cross-linking of two molecules of 2,4-diaminotoluene with
formaldehyde to give bis(2,4diamino-5-tolyl)methane.
Köppel et al. (1991) described the formation of ⫹12-Da products
from the gas chromatographic analyses of several drugs by addition of
formaldehyde and subsequent loss of water. It was suggested that the
formaldehyde was formed by thermal degradation of MeOH in the gas
chromatographic source.
These reported cases describe an analytical or in vitro artifact
from the use of MeOH as a diluent or solvent, which is then
believed to be oxidized chemically or metabolically to formaldehyde. In this study AZX ⫹ 13 (the formaldehyde addition product)
was derived directly from an in vivo investigation, suggesting a
potentially different (i.e., natural) source of formaldehyde for the
reaction. AZX ⫹ 13 was not detected in any preliminary in vitro
experiments because test compounds are typically spiked into
incubations using ACN and not MeOH. However, we have also
demonstrated that the MeOH used as mobile phase constituent
and/or indeed as the AZX diluent (used in the stock and test
solutions) did not cause the formation of AZX ⫹ 13.
Further investigations were performed to rule out MeOH as the
source of the formaldehyde or indeed as the reagent itself. The
experiments are not described in detail here; however, a brief summary is provided.
On spiking of compound AZX/MeOH into “freshly collected”
predose/control rat urine, using the MeOH/H2O gradient, AZX ⫹ 13
was produced spontaneously with 90 to 100% yield (based on the UV
response) and remained constant over 1 month (with the urine stored
at 4°C). However, on spiking of AZX/MeOH into “old” predose/
control rat urine, there was very little evidence of the formation of the
AZX ⫹ 13 product. Here, the urine was stored cold (at 4°C) for
approximately 2 months before spiking, allowing degradation/evaporation of the formaldehyde, on the basis of the half-life of formaldehyde (Organization for Economic Cooperation and Development
Screening Information Data Set, Formaldehyde, 2002, http://
www.inchem.org/documents/sids/sids/FORMALDEHYDE.pdf). After the 2-month aging period, upon spiking of AZX, only approximately 5 to 10% of the product was formed, the ratio remaining
constant over a further month.
Additional evidence on the formation of the in vivo AZX ⫹ 13 was
provided by spiking fresh rat urine samples with AZX, as conducted
in the experiments detailed above, using ACN as the spiking
solvent. The UPLC-UV-MS/MS analysis was also performed in the
Downloaded from dmd.aspetjournals.org at ASPET Journals on June 17, 2017
1
8,8’
O
REACTION OF HOMOPIPERAZINE WITH FORMALDEHYDE
O
O
R1
1
O
4
R1
8
2
+
N a,b N
5
1485
3
KCN
9
8
c,d
NC
R1
8
1 2
N
N
3
1
9
8
or
(isomer 1)
2 CN
+
N a,b N
(isomer 2)
FIG. 7. 1H NMR selective ROESY experiment
(bottom) verifying the correct structure of the
AZX ⫹ 13 ⫹ CN adduct as isomer 1 (top) after
addition of KCN.
Here, the Schiff base reactive intermediate, trapped because of the
proximity of the two nitrogens (as in a twisted boat conformation),
formed a very stable bridged homopiperazine detectable by LC-UVMS. However, in most other basic xenobiotics the Schiff base intermediate would not be intramolecularly trapped to form a stable
product detectable by subsequent analyses. Hence, there is potential
for this reactive intermediate to be missed in in vivo/in vitro metabolite identification studies, even when formaldehyde is present.
Because the initial terminal piperazine series was found to generate
electrophilic, reactive intermediates metabolically, the homopiperazine series was thought to provide a safer suitable alternative. Therefore, the formation of a bridged homopiperazine through a reactive
quaternary Schiff base intermediate was an unexpected observation.
The homopiperazine ring is neither planar nor symmetrical, so the
conformation adopted clearly proved favorable for the formation of a
one-carbon bridge.
Various other homopiperazine analogs of compound AZX have
also been investigated for ⫹13-Da product formation and as demonstrated in this study with compound AZX, these products have been
formed rapidly in vivo, as well as being confirmed in spiking experiments with formaldehyde (data not shown). These studies have
highlighted the fact that the homopiperazine moiety appears to be a
highly reactive group, and the observations made in this study support
the decision to change the chemistry of these compounds.
In conclusion, this study has demonstrated that compound AZX and
various analogs containing the homopiperazine moiety reacted rapidly
with formaldehyde present in biological fluids, as a constituent of the
one-carbon pool, suggesting that it is a metabolite, as well as a
chemical product. The homopiperazine moiety was initially included
in the lead compounds as an inert/unreactive alternative to the terminal piperazine, which was shown to be prone to bioactivation and
subsequent covalent binding with GSH. Hence, the addition of the
terminal homopiperazine group was believed to retain the potency of
the compounds while decreasing their reactivity. However, as we have
shown in this study, this group is itself highly reactive with endogenous formaldehyde, forming bridged homopiperazines.
The findings in this article suggests that compound AZX and its
homopiperazine analogs scavenge endogenous formaldehyde in vivo
similar to aminoguanidine, which was reported by Kazachkov et al.
Downloaded from dmd.aspetjournals.org at ASPET Journals on June 17, 2017
absence of MeOH, using ACN (containing 0.1% formic acid) as
the mobile phase, revealing that AZX ⫹ 13 was still formed as the
major product (between 90 and 100%), as assessed by UV quantification and MS analysis (data not shown). However, the AZX
standards (the stock and test solutions, prepared in ACN) did not
produce the product when subjected to UPLC-UV-MS/MS analysis, ruling out ACN and formate as possible reagents.
These simple experiments suggested not only that the formaldehyde
in the urine was naturally present but also that it was depleted over
time. In addition, they proved that AZX ⫹ 13 was not likely to be an
analytical artifact from the MeOH, ACN, or formate used in the
analyses.
The findings from our research appear to support the evidence in
the literature, namely that the formaldehyde occurs naturally in all
mammalian tissues, cells, and bodily fluids and that it is present in rat
blood at approximately. 0.1 mM concentration (Heck and Casanova,
2004). Formaldehyde and its oxidation product formate are reported to
be key intermediates in the “one-carbon pool” (Neuberger, 1981).
Certain xenobiotics such as MeOH, N-, O-, or S-Me compounds, and
methylene chloride apparently can also contribute to this pool. The
one-carbon pool is used for the biosynthesis of purines, thymidine,
and certain amino acids, which are incorporated into the DNA, RNA,
and proteins during macromolecular synthesis. It is therefore suggested that compound AZX scavenges endogenous formaldehyde in
vivo, similar to aminoguanidine, which was reported by Kazachkov et
al. (2007).
Formaldehyde is known to be responsible for cross-linking proteins
(e.g., Metz et al., 2004; Toews et al., 2008) and some xenobiotics
(Cunningham et al., 1990) through a Schiff base intermediate. Considering that basic amine groups are ubiquitous on many xenobiotics
and proteins, it is worth noting the potential for formaldehyde to cause
cross-linking between xenobiotics and proteins in vivo. Hence, susceptibility to formaldehyde addition via a direct or metabolic route
could lead to the formation of covalently bound protein adducts.
However, the reaction in vitro relies on the presence of MeOH, e.g.,
used as spiking solvent, as a source of the formaldehyde [as outlined
by Yin et al. (2001) and Cunningham et al. (1990)]. Therefore, the
potential for some basic xenobiotics to covalently bind to protein may
not be flagged by reactive metabolite screening or radiolabeled in
vitro covalent binding studies.
1486
MARTIN ET AL.
(2007). What appears to be a desired effect with aminoguanidine
proved to be a toxicity alert with the AZX-derived compounds.
Apart from the reactivity aspect itself, quantitation of parent drug in
blood and urine would be greatly compromised, creating problems for
lead optimization and pharmacokinetic assessments. In addition, the
formation of this adduct could artificially distort metabolite profiles in
human metabolism studies, affecting (metabolites in safety testing)
investigations (U.S. Food and Drug Administration guidance on Safety
Testing of Drug Metabolites, 2008, http://www.fda.gov/downloads/Drugs/
GuidanceComplianceRegulatoryInformation/Guidances/ucm079266.pdf).
However, overall the study demonstrated that the bridged homopiperazine was readily formed in vivo, producing a stable product/
metabolite, which did not degrade in urine for at least 1 month. This
finding has provided a suitable method and an easy and highly
efficient means (⬎95% yield) for the generation of a bridged homopiperazine synthetically by a simple reaction with formaldehyde.
Authorship Contributions
References
Argoti D, Liang L, Conteh A, Chen L, Bershas D, Yu CP, Vouros P, and Yang E (2005) Cyanide
trapping of iminium ion reactive intermediates followed by detection and structure identification using liquid chromatography-tandem mass spectrometry (LC-MS/MS). Chem Res
Toxicol 18:1537–1544.
Baillie TA (2006) Future of toxicology-metabolic activation and drug design: challenges and
opportunities in chemical toxicology. Chem Res Toxicol 19:889 – 893.
Address correspondence to: Scott Martin, DMPK Department, Alderley Park,
AstraZeneca UK Ltd., Macclesfield, Cheshire SK10 4TG, UK. E-mail:
[email protected]
Downloaded from dmd.aspetjournals.org at ASPET Journals on June 17, 2017
Participated in research design: Martin, Lenz, Temesi, Wild, and Clench.
Conducted experiments: Martin and Lenz.
Contributed new reagents or analytic tools: Martin and Lenz.
Performed data analysis: Martin and Lenz.
Wrote or contributed to the writing of the manuscript: Martin, Lenz,
Temesi, Wild, and Clench.
Baillie TA (2009) Approaches to the assessment of stable and chemically reactive drug metabolites in early clinical trials. Chem Res Toxicol 22:263–266.
Cunningham ML, Matthews HB, and Burka LT (1990) Activation of methanol by hepatic
postmitochondrial supernatant: formation of a condensation product with 2,4-diaminotoluene.
Chem Res Toxicol 3:157–161.
Doss GA, Miller RR, Zhang Z, Teffera Y, Nargund RP, Palucki B, Park MK, Tang YS, Evans
DC, Baillie TA, et al. (2005) Metabolic activation of a 1,3-disubstituted piperazine derivative:
evidence for a novel ring contraction to an imidazoline. Chem Res Toxicol 18:271–276.
Evans DC, Watt AP, Nicoll-Griffith DA, and Baillie TA (2004) Drug-protein adducts: an
industry perspective on minimizing the potential for drug bioactivation in drug discovery and
development. Chem Res Toxicol 17:3–16.
Heck H and Casanova M (2004) The implausibility of leukemia induction by formaldehyde: a
critical review of the biological evidence on distant-site toxicity. Regul Toxicol Pharmacol
40:92–106.
Kazachkov M, Chen K, Babiy S, and Yu PH (2007) Evidence for in vivo scavenging by
aminoguanidine of formaldehyde produced via semicarbazide-sensitive amine oxidasemediated deamination. J Pharmacol Exp Ther 322:1201–1207.
Köppel C, Tenczer J, and Peixoto-Menezes KM (1991) Formation of formaldehyde adducts from
various drugs by use of methanol in a toxicological screening procedure with gas chromatography-mass spectrometry. J Chromatogr 563:73– 81.
Metz B, Kersten GF, Hoogerhout P, Brugghe HF, Timmermans HA, de Jong A, Meiring H, ten
Hove J, Hennink WE, Crommelin DJ, et al. (2004) Identification of formaldehyde-induced
modifications in proteins: reactions with model peptides. J Biol Chem 279:6235– 6243.
Neuberger A (1981) The metabolism of glycine and serine, in Comprehensive Biochemistry, vol.
19A (Neuberger A and van Deenen LLM eds) pp. 257–303, Elsevier, Amsterdam.
Prakash C, Sharma R, Gleave M, and Nedderman A (2008) In vitro screening techniques for
reactive metabolites for minimizing bioactivation potential in drug discovery. Curr Drug
Metab 9:952–964.
Teschke R, Hasumura Y, and Lieber CS (1974) NADPH-dependent oxidation of methanol,
ethanol, propanol and butanol by hepatic microsomes. Biochem Biophys Res Commun 60:
851– 857.
Toews J, Rogalski JC, Clark TJ, and Kast J (2008) Mass spectrometric identification of
formaldehyde-induced peptide modifications under in vivo protein cross-linking conditions.
Anal Chim Acta 618:168 –183.
Yin H, Tran P, Greenberg GE, and Fischer V (2001) Methanol solvent may cause increased
apparent metabolic instability in in vitro assays. Drug Metab Dispos 29:185–193.